I am fairly sure that the answer to that is the FET opamp will deliver worse numbers. The lowest noise FET opamp is probably OPA627 which is still more than 10dB noisier than OPA1612.
Interesting that you mention AD811, I modified my PhiDAC (the one using AD8017) replacing those CFBs with AD811s in the I/V stage. The loss of air I heard when removing the CLC filter wasn't very much apparent when the I/V opamp was AD811. So AD811 is doing something subjectively right that AD8017 is not when they're both exposed to the raw, unfiltered DAC output.
You posted while I was writing... 
AD811 is difficult to implement (loves to oscillate - the Cf can not be used + even the attention to Rf resistor type is important) - the main reason why it is overlooked, routinely. But, with 1704 there was simply no other way to ensure near 0 dynamic impedance (which was my primary objective -> goes to show that the choice of an analog stage depends solely on the DAC specs/capabilities).
DDDAC 1974 (8 DAC's in parallel, per each channel - decent current capability) worked absolutely great with a 33-ohm resistor only; 1704 sucked big time with a resistor.
I think May uses OPA1611 as I/V followed by a class A transistor buffer; will have a look inside as soon as I get some spare time.
AD811 is difficult to implement (loves to oscillate - the Cf can not be used + even the attention to Rf resistor type is important) - the main reason why it is overlooked, routinely. But, with 1704 there was simply no other way to ensure near 0 dynamic impedance (which was my primary objective -> goes to show that the choice of an analog stage depends solely on the DAC specs/capabilities).
DDDAC 1974 (8 DAC's in parallel, per each channel - decent current capability) worked absolutely great with a 33-ohm resistor only; 1704 sucked big time with a resistor.
I think May uses OPA1611 as I/V followed by a class A transistor buffer; will have a look inside as soon as I get some spare time.
Last edited:
You are probably right. For example, a typical 8 nV/sqrt(Hz) op-amp like the OPA134 would have an A-weighted noise of about 0.9 uV when working at a unity noise gain, which is -126.9 dB(A) compared to 2 V RMS. It will be worsened by the noise gain when the output impedance of the DAC is not much greater than the feedback resistance. That would be utterly unacceptable for a 132 dB DAC like the PCM1794, and utterly negligible for a 16 bit DAC.
Last edited:
I looked at the Holo May NOS/OS Dac that was mentioned a few postings back, and a read a Stereophile review.
Although results are very impressive, I have a few things that I don't get.
1) A NOS Dac is supposed to produce a perfect impulse response, but the Holo May does not, see image 1
2) The reproduction of a 1Khz sinewave in NOS should show a perfect staircase, but the vertical lines are skewed, see fig 2
3) The Holo May stops reproducing in NOS at 20Khz, no matter whether the input is 44.4, 96 or 192Khz.
With these 3 points, I can only assume that some digital filtering is taking place, since the noise spectrum in NOS mode shows a perfect Sinc, indicating no analogue filtering ?
Last but not least the 24bit bit reproduction of an undithered -90.31dBfs input signal.
This is the best I have ever seen. Compare it to same signal processed in my heavily modified DAC with battery supply, resp. fig 3 and fig 4.
Hans
.
Although results are very impressive, I have a few things that I don't get.
1) A NOS Dac is supposed to produce a perfect impulse response, but the Holo May does not, see image 1
2) The reproduction of a 1Khz sinewave in NOS should show a perfect staircase, but the vertical lines are skewed, see fig 2
3) The Holo May stops reproducing in NOS at 20Khz, no matter whether the input is 44.4, 96 or 192Khz.
With these 3 points, I can only assume that some digital filtering is taking place, since the noise spectrum in NOS mode shows a perfect Sinc, indicating no analogue filtering ?
Last but not least the 24bit bit reproduction of an undithered -90.31dBfs input signal.
This is the best I have ever seen. Compare it to same signal processed in my heavily modified DAC with battery supply, resp. fig 3 and fig 4.
Hans
.
Attachments
Marcel & Lampie,
Could you please read the review
HoloAudio May (Level 3) D/A processor | Stereophile.com
Maybe you can get the answers.
Hans
Could you please read the review
HoloAudio May (Level 3) D/A processor | Stereophile.com
Maybe you can get the answers.
Hans
My DAC produces a measured 2.7uVrms output over 20Khz BW.
This translates into 19nV/rtHz.
With two LT1468 producing 7nv/rtHz and one OPA1632 producing ca 5nV/rtHz including 1K feedback resistors, total noise contribution is 9nV/rtHz, leaving some 17nV/rtHz for PCM 1792, or 10nV/rtHz per individual output.
So yes, that means that a fast Fet amp (opa1642?) can do the job perfectly.
Hans
Fig.1 shows the May's impulse response with 44.1kHz data in NOS mode. Ignore the very small amount of symmetrical ringing before and after the single full-scale sample, which is due to the antialiasing filter of Audio Precision's A/D converter operating at a sample rate of 200kHz. This graph indicates perfect time-domain behavior due to the absence of a low-pass reconstruction filter.
Yes, that makes sense..... so better to use analog measuring equipment to measure digital stuff..
Like my 575 Transistor Curve Tracer.... no transistors (on critical places) to be found LOL....
Tektronix 575 Transistor Curve Tracer (1957)
Like my 575 Transistor Curve Tracer.... no transistors (on critical places) to be found LOL....
Tektronix 575 Transistor Curve Tracer (1957)
Last edited:
Hi, PJN, thanks, for responding to my request.
Your subjective assessmemt with NOS echoes that of many of us. The seeming oddity is that, when you look at the objective technical differences between NOS and OS, it seems that they should, essentially, sound the same. The reason(s) they do not sound the same is what we are attempting to determine, so your input is appreciated.
I do have two related questions which I hope that you are able to address:
1) Do you still hear the characteristic NOS sound with Richard's 7th order passive filter bypassed. Just naked NOS. Assuming, that your DAC's circuit or it's PCB does not make bypassing the filter possible without risk. Which it may not, per Richard.
2) Does 7th order analog filtered NOS, sound the same as, better than, or worse than NOS with the analog filters bypassed ?
Those are key determinations for us. Thanks!
Last edited:
Thanks, Paul.
I had suspected, from Richards prior comments to me, that you might not be able to bypass the filter.
Richard indicated the same to me, about the filter order. Paraphrasing - That his 7th order filter sounded different from a 3rd order, but that a 9th order filter didn't offer a subjective improvement over the 7th order.
Interesting, how their anti-alias filter does not appear to be linear phase. Looks like a hybrid of linear and minimum phase reposnes, possibly. I've seen at least one DAC designer advocate such an reconstruction filter hybrid impulse-response.
Or, maybe, it's something else entirely.